Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. An artificial intelligence system comprising: at least one perceptual module configured to compress input received by the system to generate an internal representation that is an abstraction of features of the input received by the system; at least one cortical module configured to process the internal representation; an action selection controller configured to coordinate communication of the internal representation between the at least one perceptual module and the at least one cortical module.
2. The system of claim 1 , wherein the at least one perceptual module is further configured to decompress the internal representation.
3. The system of claim 1 , wherein the at least one perceptual module is composed of spiking artificial neurons.
The system relates to artificial neural networks designed for perceptual tasks, addressing the challenge of efficiently processing sensory data while maintaining biological plausibility. Traditional artificial neural networks often lack the energy efficiency and dynamic response characteristics of biological neural systems. This system incorporates at least one perceptual module composed of spiking artificial neurons, which mimic the behavior of biological neurons by transmitting information through discrete electrical spikes. These spiking neurons enable the system to process sensory inputs, such as visual or auditory data, in a manner that closely resembles biological neural processing. The use of spiking neurons allows for event-driven computation, where information is transmitted only when necessary, reducing energy consumption compared to conventional artificial neural networks. The system may include multiple perceptual modules, each specialized for different sensory modalities or tasks, and these modules can be interconnected to form a hierarchical or distributed architecture. The spiking neurons within the perceptual module may be configured to exhibit properties such as temporal coding, synaptic plasticity, and adaptive thresholds, further enhancing the system's ability to learn and adapt to dynamic environments. This approach aims to improve the efficiency, scalability, and biological realism of artificial neural networks for perceptual applications.
4. The system of claim 1 , wherein the at least one perceptual module is composed of non-spiking artificial neurons.
The system relates to artificial neural networks designed for perceptual processing, addressing the challenge of efficiently modeling biological neural systems in artificial intelligence. Traditional spiking neural networks mimic biological neurons by transmitting discrete spikes, but they require complex hardware and high computational resources. This system replaces spiking neurons with non-spiking artificial neurons, which use continuous activation functions like sigmoids or rectified linear units (ReLUs). These neurons process input signals through weighted connections and activation functions, enabling efficient learning and inference without the overhead of spike-based communication. The system integrates these non-spiking neurons into perceptual modules, which may include vision, audio, or sensor processing units. These modules extract features from raw data, such as identifying edges in images or detecting patterns in sound waves. The non-spiking approach simplifies implementation while maintaining performance, making the system suitable for real-time applications like robotics, autonomous vehicles, or medical diagnostics. The use of non-spiking neurons reduces hardware complexity and power consumption compared to spiking alternatives, while still supporting deep learning architectures for complex perceptual tasks.
5. The system of claim 1 , wherein the at least one cortical module is composed of spiking artificial neurons.
The invention relates to a neural network system designed for efficient and biologically plausible computation. The system addresses the limitations of traditional artificial neural networks by incorporating cortical modules that mimic the structure and function of biological neural networks. These modules are composed of spiking artificial neurons, which communicate through discrete, time-based electrical pulses similar to biological neurons. This approach enables the system to process information in a more energy-efficient and biologically realistic manner compared to conventional artificial neural networks that rely on continuous activation values. The spiking neurons allow for temporal coding, where the timing of spikes carries information, enhancing the system's ability to model complex cognitive processes. The system is particularly useful in applications requiring real-time processing, adaptive learning, and low-power computation, such as robotics, brain-machine interfaces, and autonomous systems. The use of spiking neurons also facilitates the implementation of synaptic plasticity mechanisms, enabling the system to learn and adapt dynamically to new inputs. This architecture improves computational efficiency and scalability while maintaining biological plausibility, making it suitable for advanced machine learning and artificial intelligence applications.
6. The system of claim 1 , wherein the at least one cortical module is composed of non-spiking artificial neurons.
7. The system of claim 1 , wherein the action selection controller is composed of spiking artificial neurons.
The system relates to artificial intelligence, specifically to action selection mechanisms in autonomous systems or robotic controllers. The problem addressed is the need for efficient, biologically plausible decision-making in artificial agents, where traditional artificial neural networks may lack the temporal dynamics and energy efficiency of biological neural systems. The system includes an action selection controller that determines actions for an autonomous agent based on inputs from sensors or other modules. The controller is composed of spiking artificial neurons, which are computational models that mimic the behavior of biological neurons by transmitting information through discrete electrical pulses (spikes). These neurons enable the system to process temporal information and make decisions in a manner similar to biological neural networks, improving energy efficiency and adaptability. The spiking neurons in the controller receive inputs, integrate them over time, and generate output spikes when a certain threshold is reached. This mechanism allows the system to handle dynamic environments and perform real-time decision-making. The use of spiking neurons also enables the system to learn and adapt through processes like spike-timing-dependent plasticity, where the strength of connections between neurons is adjusted based on the timing of spikes. The system may further include additional components, such as sensory processing modules, memory systems, or motor control units, which interact with the action selection controller to enable autonomous behavior. The overall architecture is designed to mimic biological neural networks, providing a more efficient and scalable approach to artificial intelligence in applications like robotics, autonomous vehicles, or vi
8. The system of claim 1 , wherein the action selection controller is composed of non-spiking artificial neurons.
9. The system of claim 1 , further comprising at least one motor module configured to decompress the transformed internal representation.
10. The system of claim 9 , wherein the at least one motor module is further configured to compress the internal representation.
11. The system of claim 9 , wherein the at least one motor module is composed of spiking artificial neurons.
12. The system of claim 9 , wherein the at least one motor module is composed of non-spiking artificial neurons.
13. The system of claim 1 , wherein the internal representation is a semantic pointer representation.
14. The system of claim 1 , wherein the at least one cortical module processes the internal representation using superposition and a compression operator, wherein the compression operator is selected from the group consisting of: circular convolution, element-wise multiplication, principle component extraction, and learned operators.
Unknown
March 30, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.